Agent skill · software engineering · aradotso
openharness-agent
openharness-agent — an agent skill for coding agents.
Why this skill is useful
Provides domain-specific commands and workflows for managing multi-agent coordination and tool execution with OpenHarness that the AI wouldn't generate on its own.
What it needs
Requires node, python, uv installed locally. About 7k tokens when loaded. Last updated 2026-07-09. 70 stars on the source repository.
What this skill does
bash git clone https://github.com/HKUDS/OpenHarness.git cd OpenHarness uv sync --extra dev bash Anthropic (default) export ANTHROPICAPIKEY=$ANTHROPICAPIKEY Moonshot / Kimi (Anthropic-compatible) export ANTHROPICBASEURL=https://api.moonshot.cn/anthropic export ANTHROPICAPIKEY=$MOONSHOTAPIKEY export ANTHROPICMODEL=kimi-k2.5 Vertex-compatible gateway export ANTHROPICBASEURL=https://your-vertex-gateway/anthropic export ANTHROPICAPIKEY=$GCPAPIKEY Bedrock-compatible gateway export ANTHROPICBASEURL=https://your-bedrock-gateway export ANTHROPICAPIKEY=$AWSAPIKEY bash oh # launch interactive TUI (React+Ink terminal UI) uv run oh # without activating venv bash Single prompt, plain text output oh -p "Explain this codebase" JSON output (structured, for programmatic consumption) oh -p "List all functions in main.py" --output-format json Streaming JSON events (real-time, line-delimited) oh -p "Fix the bug in src/api.py" --output-format stream-json Pass a file as context oh -p "Review this file" --file src/main.py Resume a previous session oh --resume <session-id> Non-interactive with permission bypass (CI/automation) oh -p "Run tests and fix failures" --permission-mode auto /help — list all 54 commands /plan — show or set current task plan /commit — stage and commit changes via git /resume — resume a previous session /memory — view or edit persistent MEMORY.md /compact — manually trigger context compression /tools — list available tools /skills — list loaded skills /hooks — show registered hooks /agents — list spawned subagents /mcp — manage MCP server connections /config — view or edit current configuration /clear — clear conversation context /exit — quit the session openharness/ engine/ # Agent loop: query → stream → tool-call → loop tools/ # 43 built-in tools (file, shell, search, web, MCP) skills/ # On-demand .md skill loading plugins/ # Commands, hooks, agents, MCP servers permissions/ # Multi-level safety: path rules, command deny hooks/ # PreToolUse / PostToolUse lifecycle …
How to use it
Reference it in AdaL, Claude Code, Cursor or any coding agent — nothing to install:
@skills aradotso/openharness-agent